Active balancer circuit

[Redpacket]

Following up discussions in thesethreads:
https://secondlifestorage.com/t-QNBBM-balancers and
https://secondlifestorage.com/t-SIMPLE-active-balancing-hardware
I've had a shot at designing an active balancer.
Please note I have not (yet) built or tested this!
Aims are:
- designed for Lithium-Ion cells, may not work on LiFePo4, my own cells are LiFePo4 ;-)
- keep it as simple as possible
- transfer energy from "full" cells only
- energy is put back into pack +ve & 0V rails
- no current flows via other cells
- select inexpensive components
- select standard resistor values
- parts used are specifically low power & low voltage
- idlecurrent should be about 1-2mA
Notes:
- it should be able to pull about 8W from the cell
- efficiency is likely >80%
- transformer design might need number of turns increasing
- it will likely have a sharp "turn-on" at the bypass voltage and may need the feedback gain tuned a bit
- it may need a small capacitor close to the pack output diode
Main partsare:
MT3608 (see eBay for modules that can be adapted)
BU7242 dual CMOS op amp
TL431A 1% regulator
Circuit diagram's here:





Discussion welcome

---

There's some other similar chips to the MT3608, eg PAM2421, AP1609

A mechanism to limit maximum current might be needed.

 

[Korishan]

Why would your transformer be designed for 12/24/48V if it's a balancer? Wouldn't it be a 1:1 for pack voltage as you're shuttling power from 1 pack to the other pack? Pack being Xp design and not system voltage design

Looking further into the design:
Am I understanding correctly that you are taking Pack voltage and boosting it to System voltage and feeding it back into the whole string? As in, not going from Pack A to Pack B, but from Pack A to String. This is odd (to me at least) in the fact that some of the power you take from the Pack will come back to it as it returns to the System rails. This seems to me like a waste of efficiency and time.
Having a design where the power is shuttled from Pack A to Pack B when A is higher would be more efficient, imho.
Granted, I'm not an EE, so I may be missing something here

 

[Redpacket]

Sorry, maybe I've confused the terms. I meant for "pack" to be the system voltage of the entire battery bank.
eg 4s = 16V ish "pack voltage"
7s = 24V pack
14s = 48V pack, etc
Correct my terms & I'll update the diagrams

Re the transformer ratio, when the MOSFET switch in the chip opens, the resulting voltage pulse rises until it finds a load. The MOSFET is rated to about 28V. So the turns ratio keeps this from getting too high, as well as the transformer providing isolation up & down the string of cells.

Am I understanding correctly that you are taking Pack voltage and boosting it to System voltage and feeding it back into the whole string? As in, not going from Pack A to Pack B, but from Pack A to String.

Yes, that's it.

This is odd (to me at least) in the fact that some of the power you take from the Pack will come back to it as it returns to the System rails. This seems to me like a waste of efficiency and time.
Having a design where the power is shuttled from Pack A to Pack B when A is higher would be more efficient, imho.
Granted, I'm not an EE, so I may be missing something here

No you're not missing things, but the trick is simplicity.
The QNBBM 1s units already dynamically transfer energy straight from higher cells to lower cells but they're ~20USD each.
I was aiming for simple so this is an intermediate solution hopefully better than just dumping into resistors & cheaper than the QNBBM units.

 

[watts-on]

Interesting idea. So, all of the transformeroutputs would be connected in parallel?

 

[Korishan]

Couldn't you just raise the voltage slightly higher than the lowest voltage pack and move the power across 2 wires to the next pack? Instead of dealing with the higher voltages and needing to maintain an active voltage check on the system voltage.

Lets say there's 3 packs, for simplicity.
4.16, 4.09, 4.11 = 12.36, or 4.12 average

So, the brains would see that pack 2 is low and pack 1 is high. The FET would be driven to make the output voltage of the transformer 4.12V to push power into pack 2. Then the transformer on the pack 2 would take the 4.12V (as this is an isolation mode) and move it to the pack, and a capacitor would smooth out the ripple to make it not pwm into pack 2.
Then once Pack 2 is as at 4.12, or there abouts, the next highest pack repeats the process to the lowest pack.

That's how I envision active balancing. Seems to me a waste to boost the power all the way up to system voltage just for it to leak back down the pack your drawing power from to begin with.

 

[Korishan]

Interesting idea. So, all of the transformeroutputs would be connected in parallel?

This is a good question. I didn't see that at first. So I would assume so. This causes another issue. If there's pulsing on the system lines then there will be pulsing on the transformer, thereby pushing power backwards. Unless you have diodes and/or FETs to make sure of this. If diodes, there's losses right there. If FETs, they better be rated high enough to handle the current.

 

[watts-on]

Couldn't you just raise the voltage slightly higher than the lowest voltage pack and move the power across 2 wires to the next pack? Instead of dealing with the higher voltages and needing to maintain an active voltage check on the system voltage.

Lets say there's 3 packs, for simplicity.
4.16, 4.09, 4.11 = 12.36, or 4.12 average

So, the brains would see that pack 2 is low and pack 1 is high. The FET would be driven to make the output voltage of the transformer 4.12V to push power into pack 2. Then the transformer on the pack 2 would take the 4.12V (as this is an isolation mode) and move it to the pack, and a capacitor would smooth out the ripple to make it not pwm into pack 2.
Then once Pack 2 is as at 4.12, or there abouts, the next highest pack repeats the process to the lowest pack.

That's how I envision active balancing. Seems to me a waste to boost the power all the way up to system voltage just for it to leak back down the pack your drawing power from to begin with.

Your thinking of the perfect Active Balancer not the cheap and simple version.

RedPacket's design doesn't require any "Brains" doing any monitoring and no complicated switching to route power around in the required combination at that moment.
If my understanding is correct, you just wire it all together and it does its thing automatically.

 

[daromer]

Its doable and if you search for scientific datasheets and documents you will find fullproof tests of it and systems running with above design It works for sure and was one design I looked at 2.5 year back when I started investigate build my own system

 

[Redpacket]

Interesting idea. So, all of the transformeroutputs would be connected in parallel?

This is a good question. I didn't see that at first. So I would assume so. This causes another issue. If there's pulsing on the system lines then there will be pulsing on the transformer, thereby pushing power backwards. Unless you have diodes and/or FETs to make sure of this. If diodes, there's losses right there. If FETs, they better be rated high enough to handle the current.

Yes allthe transformer outputs would be in parallel after the diode. As per the diagram, yes there's a diode there. No power would flow backwards from the system voltage rail because of the diode.
The principle at work here is the same as a "flyback" power supplydesign.

Re losses, our starting point is dumping 100% of energy into resistors & just wasting all ofit - so adiode's loss isn't the end of the world!
If you did want to step upthe complexity (& cost), a synchronous rectifier could be added at the transformer output. MOSFETs can easily handle a few amps!

---

Couldn't you just raise the voltage slightly higher than the lowest voltage pack and move the power across 2 wires to the next pack? Instead of dealing with the higher voltages and needing to maintain an active voltage check on the system voltage.

Lets say there's 3 packs, for simplicity.
4.16, 4.09, 4.11 = 12.36, or 4.12 average

So, the brains would see that pack 2 is low and pack 1 is high. The FET would be driven to make the output voltage of the transformer 4.12V to push power into pack 2. Then the transformer on the pack 2 would take the 4.12V (as this is an isolation mode) and move it to the pack, and a capacitor would smooth out the ripple to make it not pwm into pack 2.
Then once Pack 2 is as at 4.12, or there abouts, the next highest pack repeats the process to the lowest pack.

That's how I envision active balancing. Seems to me a waste to boost the power all the way up to system voltage just for it to leak back down the pack your drawing power from to begin with.

I thought about this too - problem is thatif adjacent cells in a string were designed to dumpexcesspower to the next cell (either up or down) in the string, you'd easily get a situation where say 3 in a row high cells are concentrating too much current to balance.

If you don't push the excess energy to the battery system rail & instead collect it close to cell voltage (but not adjacent cell either), then you need another bunch of isolated converters (one per cell) to push it into low cells.
So there goes the complexity/cost up again!

One way (in theory, but way too impractical!!!) could be that each of the transformers in the design above has many duplicate windings, eg one output winding per cell in the string.
Each balancer's multiple outputs would be across multiple cells - feel the wiring nightmare coming already?
Might just be doable for a smaller "12V" few cell system but a 24V 7s system would need 7x windings, a 48V system 14x windings (no chance!).

There's no easy 100% win-win ;-)

 

[daromer]

If you want to dump load into the lowest cell you need to look into Biderctional DC/DC instead but as you stated its more complexity

 

[Redpacket]

Its doable and if you search for scientific datasheets and documents you will find fullproof tests of it and systems running with above design It works for sure and was one design I looked at 2.5 year back when I started investigate build my own system

Hi Daromer, can you remember the link to this design?

 

[Cherry67]

Looks like i have awakened interest....

I am working on a design as well, but MUCH cheaper.
Less Power. But i have to check it first....

 

[Bubba]

Would this be doable? I am new to arduino so not sure if this could be pulled off.
I read here -->http://www.home-automation-communit...n-atmega328p-for-a-year-on-coin-cell-battery/
About low power Arduino usage.

The new Smart BMS's from china have UART outputs.
Could use the output to switch a mosfet connecting a DC-DC buck converter from full pack voltage to the lowest cell voltage.

atmega328 pro mini which is low power could be used to UART connect to the bms and the 14 digital arduino pins could control the mosfets?

 

[Cherry67]

Bubba, interesting, but powering the Arduino down is not the major problem.

 

[Redpacket]

Looks like i have awakened interest....

I am working on a design as well, but MUCH cheaper.
Less Power. But i have to check it first....

Can you share what you'e thinking of?

 

[Cherry67]

Looks like i have awakened interest....

I am working on a design as well, but MUCH cheaper.
Less Power. But i have to check it first....

Can you share what you'e thinking of?

Yes. Before or after the first Test?

 

[Redpacket]

Yes. Before or after the first Test?

Any time, so we can discuss!

---

PS,
I've built a "proof of concept" version of the circuit way above - with a few tweaks and using some substitute parts I had in hand, it's working "stage 1".
I'm testing it on a 3s/2p laptop battery I found. The battery has good Samsung pink 26F cells & seems healthy after the built in BMS was removed.

Please note right now I'm using a current limited "lab" supply as the source (don't want smoke in testing phase!). So far it starts pushing current back into the main battery rail (~11.4V now), starts at ~4.1V and with the parts used so far is transferring about 1.8W. There's a few more tweaks needed yet.

One of the questions here is - "is it worth it?" In the end we're only going to recover a few watts per cell x number of cells (although design can be scaled up), so at the small end, this might be say 2W/cell x 7 for a 24V system with say 6/7 bypassing = 12W for part of the day. Not really very much power recovered.
Even in a larger system at say 20W/cell, this is only 120W for a few hours.

---

Another chip type has occurred to me: LED driver chips. These might be adapted to this as well. Eg a ZXSC400 with an external FET.

 

[Cherry67]

What i had in mind is something totally different.
about 20 mA only, active balancing as "low balancing", means if a cell gets lower as the others starting from - say 3,7 Volt.

The minimal solution i had in mind is too simple, i think.... but the basic concept might still be interesting.